The useful distinction is temporal: can the named person fail the current run before it ships, or only change what happens next? This page uses human in the loop for the first and human on the loop for the second.
Human on the loop, on this site, is a named monitor who can stop the next run after watching this one. Human in the loop is a named person who can fail this run before it ships. Human in the loop meaning owns the in-the-loop definition. This page owns the current-run versus next-run distinction.
This split is house usage. It is not an IEEE standard. If a vendor slide uses the phrases the other way, ask which run the person can actually fail.
The reach test
Stand in the room and ask: can this person stop the sheet that is printing now?
If yes, with written criteria and a name, you have in-the-loop. If they can only pull a lever after the sheet is stacked, you have on-the-loop. If nobody is named, you have neither.
Classify the role by intervention timing, not by the dashboard it watches. If the current output can ship before that person acts, the person is not in this run’s gate; they may still govern the next run.
Collision control
Not a synonym. HITL on this site is in-the-loop. On-the-loop is the weaker monitor.
Not override doctrine. Override doctrine is what the agent may never do. The loop is who can stop it. Different objects.
Not the agent. What is an AI agent? owns the worker.
What this page is not
It is not a standards brief. It is not a vendor glossary. It is not the full HITL craft.
Write which run the named person can fail. If the answer is “the next one,” you are on the loop. Stop calling that in the loop.
Cite this:Human on the loop: watching is not a fail gate.Len P. van der Hof. https://lenvanderhof.com/en/blog/human-on-the-loop/ ·